Bug 14726 - [845G] X Server fails to start with "intel" driver, works with "i810".
Summary: [845G] X Server fails to start with "intel" driver, works with "i810".
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: 7.2 (2007.02)
Hardware: x86 (IA32) Linux (All)
: high major
Assignee: Hong Liu
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-02-28 15:19 UTC by Vincent René-Corail
Modified: 2008-03-19 19:47 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
This is how I have changed it. (2.48 KB, text/plain)
2008-03-06 14:26 UTC, Vincent René-Corail
no flags Details
Xorg.0.log with "i810" driver enabled. (53.96 KB, text/plain)
2008-03-07 15:12 UTC, Vincent René-Corail
no flags Details
Xorg.0.log with "intel" driver enabled. (60.70 KB, text/plain)
2008-03-07 15:26 UTC, Vincent René-Corail
no flags Details
Xorg.0.log_i810_driver (54.00 KB, text/plain)
2008-03-10 04:55 UTC, Vincent René-Corail
no flags Details
Xorg.0.log_intel_driver (78.80 KB, text/plain)
2008-03-10 04:56 UTC, Vincent René-Corail
no flags Details
Q910_screen_specifications (231.74 KB, image/png)
2008-03-11 13:38 UTC, Vincent René-Corail
no flags Details
recommended_resolution_and_refresh_rate (169.50 KB, image/png)
2008-03-11 13:58 UTC, Vincent René-Corail
no flags Details
Newly edited xorg.conf with "intel" driver enabled (2.74 KB, text/plain)
2008-03-13 02:40 UTC, Vincent René-Corail
no flags Details
xorg.0.log with working xorg.conf edited with "intel" driver (125.38 KB, text/plain)
2008-03-13 02:50 UTC, Vincent René-Corail
no flags Details
xorg.conf from Mandriva 2008 liveCD (2.65 KB, text/plain)
2008-03-17 15:43 UTC, Vincent René-Corail
no flags Details
Xorg.0.log from Mandriva 2008 liveCD (with debug mode) (53.77 KB, text/plain)
2008-03-17 15:45 UTC, Vincent René-Corail
no flags Details

Description Vincent René-Corail 2008-02-28 15:19:38 UTC
Hi,

I've tried to sort everything out according to this page http://intellinuxgraphics.org/how_to_report_bug.html, I hope my report is clear.

I reported that bug in launchpad (Ubuntu), and I've been told to report it here, so, here I am.

There are all the explanations :

https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/191000

Feel free to ask anything that might help you.

Vincent
Comment 1 Gordon Jin 2008-02-28 18:47:32 UTC
We don't have the "common sense" which driver version included in Ubuntu. You nned provide the full info here, with the template on http://intellinuxgraphics.org/how_to_report_bug.html.
Comment 2 Hong Liu 2008-02-28 19:00:14 UTC
Please make sure modedebug is turned on when attaching the xorg.log
Comment 3 Vincent René-Corail 2008-02-29 13:51:42 UTC
Ok, I'll try to do my best, but please, remember I'm confirmed user but not an expert (nor a developper) and moreover, I'm not naturally english speaking.
I'm just saying that to say that it's diffucult to understand and then provide all the informations (very technical language, etc.)... But I want to do it well.

First, the hardware :

http://launchpadlibrarian.net/12264848/lspci-vvnn_output

845G

-- uname -m = i686

-- kernel version = 2.6.22-14-generic

-- Linux distribution : Ubuntu 7.10 (seems to be exactly the same bug in future 8.04 at least until alpha 5... exactly the same symptoms)

-- Machine or mobo model : Old Nec machine I got recently (2.2 GHz celeron processor, integrated intel graphics card, etc.)

3) Reproduce steps. Probability if not 100% reproducible :

Just take an Ubuntu 7.10 (final) Live CD (Desktop CD) or Ubuntu 8.04 (until alpha 5 at least) and the symptoms should be the same (X server doesn't start) with the same kind of hardware... I think so.

I can't go further for tonight (late and working tomorrow), I'll come back tomorrow or in the two next days to provide the attachments asked there http://intellinuxgraphics.org/how_to_report_bug.html in debug mode.

Before that, I need some explanations about the third point of the "how_to_report_bug" page.

My situation seems to be the same as described :

"-- If the display mode broken with the current driver but you have another method to get it work (e.g. using the old i810-1.x driver), use reg_dump tool (in xf86-video-intel package) to dump the register info under the both conditions."

I am not sure I have understood what the tool to "dump" the "register" is and how to activate it.

It's something to clear or "make empty" the "register" ("register" between brackets because I don't know nor understand what it is) ?

If yes, how do I activated it ?

See you in a few days.

Vincent
Comment 4 Hong Liu 2008-03-03 01:26:25 UTC
(In reply to comment #3)
> I am not sure I have understood what the tool to "dump" the "register" is and
> how to activate it.
> 

It's just a tool in xf86-video-intel/src/reg_dump, when you compile the video driver, the tool "intel_reg_dump" will be compiled too. Dumping just mean reading the content of the registers by using this tool.

Thanks,
Hong
Comment 5 Vincent René-Corail 2008-03-06 14:26:58 UTC
Created attachment 14898 [details]
This is how I have changed it.
Comment 6 Vincent René-Corail 2008-03-06 14:30:10 UTC
Comment on attachment 14898 [details]
This is how I have changed it.

I've tried to activate the debug mode, but everytime I change my xorg.conf file (like this : "3. Assistance to help debugging:
-- Add option "ModeDebug" "yes" in the Device section of xorg.conf to dump more register information into Xorg log.") and reboot, x server does not start.
Could you help me activated the debug mode correctly please ?
Comment 7 unggnu 2008-03-06 17:35:31 UTC
I guess you have to replace the "yes" with "true".
Comment 8 Vincent René-Corail 2008-03-07 11:41:17 UTC
(In reply to comment #7)
> I guess you have to replace the "yes" with "true".
> 

I think I've tried everything. In the "Section Device" part, I've tried to add :

ModeDebug "yes"

ModeDebug "true"

ModeDebug yes

ModeDebug true

Modedebug "yes"

Modedebug "true"

Modedebug yes

Modedebug true

And same thing with "DebugMode" (instead of ModeDebug).

... But no way, it won't work. Even with the driver set to "i810", which works normally when I don't try to enable the debug mode.
Comment 9 Daniel Stone 2008-03-07 13:59:02 UTC
On Fri, Mar 07, 2008 at 11:41:18AM -0800, bugzilla-daemon@freedesktop.org wrote:
> > I guess you have to replace the "yes" with "true".
> 
> I think I've tried everything. In the "Section Device" part, I've tried to add
> :
> 
> ModeDebug "yes"

You need:
Option "ModeDebug"
instead of just:
ModeDebug
Comment 10 Vincent René-Corail 2008-03-07 15:09:41 UTC
(In reply to comment #9)
> On Fri, Mar 07, 2008 at 11:41:18AM -0800, bugzilla-daemon@freedesktop.org
> wrote:
> > > I guess you have to replace the "yes" with "true".
> > 
> > I think I've tried everything. In the "Section Device" part, I've tried to add
> > :
> > 
> > ModeDebug "yes"
> 
> You need:
> Option "ModeDebug"
> instead of just:
> ModeDebug
> 

FINALLY !!! It finally works, so I'll attach the two Xorg.0.log in a few minutes (first with "i810" and second with "intel").
Comment 11 Vincent René-Corail 2008-03-07 15:12:44 UTC
Created attachment 14947 [details]
Xorg.0.log with "i810" driver enabled.
Comment 12 Vincent René-Corail 2008-03-07 15:26:50 UTC
Created attachment 14948 [details]
Xorg.0.log with "intel" driver enabled.
Comment 13 Hong Liu 2008-03-09 20:31:26 UTC
(In reply to comment #12)
> Created an attachment (id=14948) [details]
> Xorg.0.log with "intel" driver enabled.
> 

I can't see your EDID data, would you please add

Option "ModeDebug" "true" to your device section and repost the xorg.log?

Thanks,
Hong
Comment 14 Vincent René-Corail 2008-03-10 04:55:24 UTC
Created attachment 14999 [details]
Xorg.0.log_i810_driver
Comment 15 Vincent René-Corail 2008-03-10 04:56:15 UTC
Created attachment 15000 [details]
Xorg.0.log_intel_driver
Comment 16 Hong Liu 2008-03-10 18:53:35 UTC
What's the native resolution of your monitor?

From the log:
(II) intel(0): First detailed timing not preferred mode in violation of standard!

It seems there are some problems in your EDID data block, which causes the driver to select the unsupported 1600x1024 as the default mode.

Thanks,
Hong
Comment 17 Vincent René-Corail 2008-03-11 13:29:37 UTC
My screen is a Hyundai ImageQuest Q910 (CRT). Its native resolution is 1600x1200@85Hz or 1280x1024@85Hz (I always set it at 1280x1024@85Hz otherwise, everything is too small).
Comment 18 Vincent René-Corail 2008-03-11 13:38:54 UTC
Created attachment 15039 [details]
Q910_screen_specifications
Comment 19 Vincent René-Corail 2008-03-11 13:58:03 UTC
Created attachment 15040 [details]
recommended_resolution_and_refresh_rate

Don't hesitate if you need anything else. I'll try to answer as soon as possible.

Thanks.

Vincent

PS : I don't know if you can see it with the logs I have sent you, but even if I have set (manually in the xorg.conf file) Horizontal : 30 to 107kHz and Vertical : 50 to 150Hz, and in the gnome settings 1280x1024@85Hz, the horizontal frequency is 93.3 KHz and vertical at 87 Hz, instead of 91.1 KHz and 85 Hz as recommended by the manufacturer (Mandriva Linux was detecting and configuring it correctly, not Ubuntu... I don't know if this can help).
Comment 20 Hong Liu 2008-03-11 19:35:21 UTC
(In reply to comment #19)
> PS : I don't know if you can see it with the logs I have sent you, but even if
> I have set (manually in the xorg.conf file) Horizontal : 30 to 107kHz and
> Vertical : 50 to 150Hz, and in the gnome settings 1280x1024@85Hz, the
> horizontal frequency is 93.3 KHz and vertical at 87 Hz, instead of 91.1 KHz and
> 85 Hz as recommended by the manufacturer (Mandriva Linux was detecting and
> configuring it correctly, not Ubuntu... I don't know if this can help).
> 
There are several 1280x1024@85Hz modes, it depends on which mode selected.

So does your monitor display correctly when changing to 1280x1024 mode?

If it is OK, then you may add some options to your xorg.conf to let X server selects 1280x1024@85Hz as the default mode. Since we can't get it from your EDID data.

There is a doc on how to configure the preferredmode option
http://www.intellinuxgraphics.org/dualhead.html 

You need to add:
1.   Option "Monitor-VGA" "Q910" to your device section
2.   add
   Modeline "1280x1024_OK"  157.50  1280 1344 1504 1728  1024 1025 1028 1072 +hsync +vsync
   Option "PreferredMode" "1280x1024_OK" 
to your monitor section.

Thanks,
Hong
Comment 21 Vincent René-Corail 2008-03-13 02:38:30 UTC
Here is my new xorg.conf (and log) file which I edited by mixing some part of my old Mandriva 2008 xorg.conf and what you told me.
It works perfectly well with the "intel" driver. Just one thing. Note that I couldn't enable the window effects (compiz-fusion) with screen depth set to 24 ; it would only work in 16 bit. I remember that this "bug" is already known.
Comment 22 Vincent René-Corail 2008-03-13 02:40:30 UTC
Created attachment 15081 [details]
Newly edited xorg.conf with "intel" driver enabled
Comment 23 Vincent René-Corail 2008-03-13 02:50:39 UTC
Created attachment 15082 [details]
xorg.0.log with working xorg.conf edited with "intel" driver

Note that horizontal and vertical refresh are now almost perfect. Respectively 91.5 KHz and 84 Hz (this is what my screen's osd reports). Before that (past xorg.conf with i810 driver), it was 93.3 KHz and 87 Hz (too high).

Thanks.

Vincent
Comment 24 Hong Liu 2008-03-13 17:55:55 UTC
OK, glad to hear it works.

Since we can't figure out which mode is suitable for default from your monitor's EDID data nor can we do a quirk for it, adding preferredmode looks like the only way to work out this problem.

So I am closing it now for displaying with intel driver is no problem.

Thanks,
Hong
Comment 25 Vincent René-Corail 2008-03-15 13:47:51 UTC
(In reply to comment #24)
> OK, glad to hear it works.
> 
> Since we can't figure out which mode is suitable for default from your
> monitor's EDID data nor can we do a quirk for it, adding preferredmode looks
> like the only way to work out this problem.
> 
> So I am closing it now for displaying with intel driver is no problem.
> 
> Thanks,
> Hong
> 

Ok, I think I have understood everything.
My monitor doesn't give or send an information (EDID data) that the driver needs to set it correctly (resolution, frequency, etc.).

But the thing I still don't get is why Mandriva set everything perfectly and not Ubuntu whereas both distribution use the same driver (the same that work now after have edited my xorg.conf file) on the same hardware ?

Logically, if what you said is right, the explanation is that the problem comes from the distribution (Ubuntu), not the driver. Am I right ?

Thanks.

Vincent
Comment 26 Hong Liu 2008-03-16 17:57:01 UTC
(In reply to comment #25)

> But the thing I still don't get is why Mandriva set everything perfectly and
> not Ubuntu whereas both distribution use the same driver (the same that work
> now after have edited my xorg.conf file) on the same hardware ?
> 

Would you please attach the xorg.conf and xorg.log (with modedebug enabled) in your Mandriva?
Comment 27 Vincent René-Corail 2008-03-17 15:43:52 UTC
Created attachment 15237 [details]
xorg.conf from Mandriva 2008 liveCD

I hope it's ok if I attach the files from the liveCD (Mandriva 2008). I have added Option "ModeDebug" "yes" to the liveCD's xorg.conf file and then restarted X (ctrl + alt + back space).
Is it ok, or should I install it to the hard drive ?

Vincent
Comment 28 Vincent René-Corail 2008-03-17 15:45:03 UTC
Created attachment 15238 [details]
Xorg.0.log from Mandriva 2008 liveCD (with debug mode)
Comment 29 Hong Liu 2008-03-19 19:47:28 UTC
Hi, Vincent

Your Mandriva is using an old X server which behaves a little different with the new one in your ubuntu.

 Subsection "Display"
        Depth 24
        Modes "1280x1024" "1152x864" "1024x768" "832x624" "800x600" "640x480" "480x360" "320x240"
    EndSubsection

Old server will hornor this option and use 1280x1024 as the initial mode to set, while new server doesn't.

The EDID data from your mandriva log is the same as the one from ubuntu.

Thanks,
Hong


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.